Neuron Explorer

Qwen3-0.6B

'process' after 'is a' definitions

A neuron in layer 15 of Qwen3-0.6B. Its strongest snippets are very consistent: they mostly peak on the same token. It is featured, so it has steering outputs.

What this neuron does

This neuron lights up on the word 'process' in definitions like 'Brining is the process of...' or 'Mulching is a process that...'. Turning it up did not produce any clear change in the writing that matches this pattern.

Written by Claude from this neuron's activations and steering outputs. Confidence: medium. Think it is wrong? Flag it (opens a prefilled issue on GitHub).

Fires most on

The text windows where this neuron's activation was highest. Amber marks positive activation; the outlined token is the peak. Hover a token for its strength.

  1. #1 peak

    Project plans⏎- Project activities⏎- Legislation and standards⏎- Industry context⏎Last edited 22 Oct 2020⏎Land acquisition is the process of buying a piece of land. The motivation behind land acquisition can be varied;

  2. #2 peak

    their body gets too cold they might not be able to return to the hive .<|endoftext|>What Is Laser Scan Service in BIM?⏎Scan-to-BIM is the process of recording an existing structure to use Building Information Modeling as a digital planning tool (

  3. #3 peak

    … Also, poultry must not be left outside the refrigerator for more than two hours.⏎What does putting chicken in salt water do?⏎Brining is the process whereby chicken (or other meat) is submerged in a salty liquid (“brine

  4. #4 peak

    including:⏎· Species distribution and abundances⏎· Species movements and life stages⏎· Ecosystem processes⏎· Climate change<|endoftext|>What is Encryption?⏎Encryption is a process by which we use software to scramble sensitive information while it is in transit to Dell

  5. #5 peak

    .⏎Defuse and DiffuseThe words defuse and diffuse sound similar, but their meanings are very different.⏎DefuseThe verb to defuse refers to the act of deactivating a bomb. (Defuse literally means to remove the fuse

  6. #6 peak

    ational - Arguments / Dispute.⏎- Dissent - Strong disagreement or dissatisfaction.⏎- Fissures - Split or deep crack.⏎- Condemnation - Act of saying that something is unacceptable.⏎- Consensus - Majority of Opinion.⏎- Arch

Steering

Push this neuron up or down while the model writes, and watch what changes. The outputs were generated ahead of time, so nothing runs in your browser.

turned down a lot
Prompt

Here is something I have been thinking about lately: (I) = (A + B) + (A + B) + (A + B) + (A + B) + (A + B) + (A + B) + (A + B) + (A

Most negative on

Where the activation went furthest below zero. Blue marks negative activation. Many neurons mean something different in each direction.

  1. #1 peak

    set?⏎Click below to see some definitions used to describe graphs.⏎Can you find examples on the cards that match each definition? Can you draw some more examples of your own?<|endoftext|>Since 2012, the global economy has been experiencing

  2. #2 peak

    000 hired farmworkers are between the ages of 14 and 16. The most important of the proposed child labor changes—the first in more than three decades—that crop producers need to be aware of include:⏎- The tractor-s

  3. #3 peak

    2.5 Australia Licence together with any terms, conditions and exclusions as set out in the website Copyright notice. For permission to do anything beyond the scope of this licence and copyright terms contact us.<|endoftext|>Water scarcity is a global problem. From

  4. #4 peak

    somebody tries to cheat a victim out of money that involves sending a letter through the mail, whether the action is successful or not. The recipient of a letter with this intent can immediately report the perpetrator to the appropriate authorities. Mail fraud is a federal

Connections

Show on map

Which neurons in earlier layers feed this one, and which later ones it feeds. Amber links excite, blue links suppress; longer bars are stronger.

Fed by earlier layers

  1. L11:1338 suppresses
  2. L6:2287 suppresses
  3. L10:698 excites
  4. L3:2257 suppresses
  5. L10:998 excites
  6. L4:2379 suppresses
  7. L5:2230 excites
  8. L6:2766 excites
  9. L3:2096 excites
  10. L11:532 excites verb after a gerund subject
  11. L8:1740 weakly excites
  12. L11:1028 weakly excites
  13. L4:1936 weakly excites
  14. L5:1539 weakly excites
  15. L0:116 weakly excites
  16. L6:213 weakly suppresses
  17. L6:2322 weakly excites
  18. L7:697 weakly excites
  19. L6:2467 weakly suppresses
  20. L9:2024 weakly excites
  21. L6:964 weakly excites
  22. L5:184 weakly excites
  23. L9:1816 weakly suppresses
  24. L12:2867 weakly excites

Feeds later layers

  1. L19:1151 excites
  2. L21:520 suppresses
  3. L18:2079 excites
  4. L16:1641 excites
  5. L20:1097 excites
  6. L26:2243 excites
  7. L18:993 suppresses
  8. L22:361 excites
  9. L18:2274 suppresses
  10. L20:1168 excites
  11. L19:1307 excites
  12. L16:2270 excites
  13. L21:781 excites
  14. L22:1828 excites
  15. L21:1282 excites
  16. L19:1477 excites
  17. L18:353 excites
  18. L16:2290 suppresses
  19. L23:1467 excites
  20. L16:1360 suppresses
  21. L17:871 weakly excites
  22. L17:691 weakly excites
  23. L20:480 weakly excites
  24. L25:953 weakly excites

Activation histogram

How often this neuron fires at each strength. Almost always it does close to nothing (the tall bars in the middle); the snippets above are the rare moments it fired hard, amber for positive and blue for negative.

1 values between -2.19 and -2.001 values between -1.81 and -1.632 values between -1.63 and -1.443 values between -1.44 and -1.256 values between -1.25 and -1.0614 values between -1.06 and -0.8835 values between -0.88 and -0.6961 values between -0.69 and -0.50187 values between -0.50 and -0.32477 values between -0.32 and -0.131962 values between -0.13 and 0.06680 values between 0.06 and 0.25193 values between 0.25 and 0.4393 values between 0.43 and 0.6261 values between 0.62 and 0.8136 values between 0.81 and 0.9922 values between 0.99 and 1.1816 values between 1.18 and 1.3723 values between 1.37 and 1.558 values between 1.55 and 1.7411 values between 1.74 and 1.932 values between 1.93 and 2.122 values between 2.12 and 2.301 values between 2.30 and 2.491 values between 2.49 and 2.683 values between 2.68 and 2.862 values between 2.86 and 3.051 values between 3.05 and 3.241 values between 3.24 and 3.421 values between 3.42 and 3.611 values between 3.98 and 4.171 values between 5.11 and 5.29offnegativepositiveoften

Logit lens

The words this neuron nudges the model toward (amber) or away from (blue) when it fires. A hint, not a label: many neurons work through later layers instead.

Promotes

  1. process
  2. process
  3. processes
  4. Process
  5. 过程
  6. 的过程
  7. _process
  8. →process
  9. Process
  10. processo

Suppresses

  1. icism
  2. scarc
  3. àng
  4. legen
  5. marked
  6. Aires
  7. vature
  8. buurt
  9. 蔼
  10. markt